That's hard telecine. It both looks interlaced (2 of every 5 frames are interlaced) and is encoded as interlaced. When returned to the way they are on the DVD:
aa bb bc cd dd...
you see that 2 of the 5 frames contain images from 2 different points in time.
If the IVTC script looked OK when stepping through it in VDub, (for example), then perhaps the problem is with some setting in the encoder.
Try StreamFab Downloader and download from Netflix, Amazon, Youtube! Or Try DVDFab and copy Blu-rays! or rip iTunes movies!
+ Reply to Thread
Results 121 to 150 of 212
Thread
-
-
-
Yes, manono, that's what the encode looks like (I hope I described it sufficiently). I'm looking at TMPGenc now -- I don't have other encoders on the PC I'm using right now. I'm telling TMPGenc that the input is progressive. TMPgenc looks at the video and fills in Frame Rate as " 23.976 fps (internally 29.97 fps)". If then choose Encode Mode as "3:2 pulldown when playback", the encoded output is always interlaced with pulldown, as described above. If I want 29.97 fps playback, that's the only combination of Frame Rate and Encode Mode that TMPGenc will let me get away with if I want 29.972 fps from the clip. According to TMPgenc docs, output from that 3:2 Encode Mode is always interlaced. There seems to be no other way to get a 3:2 pulldown without interlacing the input. If I choose "progressive", the frame rate becomes 23.976, and trying to ask for 29.97 gives an error message.
OK. I'll keep looking at these crazy settings. -
I don't (and won't) use TMPGenc. But when you say, "the encoded output is always interlaced with pulldown", that's not correct if I understand what you're saying. The encoded output - what will be stored on the DVD - is progressive 23.976fps. It's the playback that's interlaced - progressive 23.976 with pulldown plays back as interlaced 29.97fps. But maybe we're using different terms to describe the same thing.
jagabo knows the proper settings to use for TMPGEnc, and if I remember what he's suggested before, what you've done is correct. As far as it goes. -
Right, manono, the playback isn't what's actually encoded, it's what is -- well, "played back", as you say. I have made progress with TMPGenc, discovered that the "interlaced" or "progressive" description of the input has to be checked again and again; that's one setting that isn't saved with other app defaults, you have to keep checking it with every new input clip. TMPGenc insisted on seeing this input as interlaced, so I changed that to progressed. So can compare new results with the source. I see TMPGenc makes some of these pulldown frames look interlaced (or just blended ??. I'll check tomorrow to see how it plays, but it does look better than the earlier encodes:
Scene with Man on bicycle (very fast motion)
Using SeparateFields()
Left-hand column = field pairs (TFF) in each frame
middle column = source fields displayed
right column = encoded fields displayed
158-159 - a/a - - a/a
160-161 - b/b - - b/b
162-163 - b/c - - bc/bc
164-165 - c/d - - bc/cd
166-167 - d/d - - cd/e
168-169 - e/e - - e/e
170-171 - f/f - - f/f
172-173 - f/g - - fg/fg
174-175 - g/h - - fg/h
176-177 - h/h - - h/h
178-179 - i/i - - i/i
180-181 - j/j - - j/jLast edited by sanlyn; 20th Mar 2012 at 07:45.
-
In the above post, I can see that the original progressive/telecined source is duplicating fields to achieve NTSC 3:2 pulldown on a 24fps > 30fps film original. In the TMPGenc example, the effect on duplicated frames is combing in the duplicated fields/frames. So looking at the encode, the original progressive frames still look progressive, but the pulldown frames look interlaced. At least, that's the way they look on the PC. I'll recode this scene, burn to DVD, and see how it plays on TV. Will try it with HCenc as well.
But a question: for every 5 frames of telecine, should there be 2 composite frames in every group of 5 instead of 3 frames in every 5?Last edited by sanlyn; 20th Mar 2012 at 08:52.
-
That's the way they're supposed to look on a PC. The editor performs the soft pulldown. The reason editors (VirtualDub, for example) do this is because the the progressive frame rate can be anywhere from 19.98 to 29.97 fps. But the final output is always 59.94 fields per second, 29.97 frames per second when packaged as frames.
You can use VirtualDubMod to view the video without performing soft pulldown. But you may not get the correct frame rate. For example, if you have a 25 FPS PAL source converted to NTSC keeping the 25 fps progressive frames and applying 3:2:3:2:2 pulldown flags, VirtualDubMod will give you all progressive frames but call it 23.976 fps rather than 25 fps.
It depends on what pulldown pattern was used. Some programs use 2:3:3:2 or 3:2:2:3 instead of 2:3 or 3:2.
2:3:3:2: AA BBB CCC DD EE FFF GGG HH -> AA BB BC CC DD EE FF FG GG HH
3:2:2:3: AAA BB CC DDD EEE FF GG HHH -> AA AB BC CD DD EE EF FG GH HH
In the 2:3:3:2 example above all the original progressive frames can be found in the 29.97i video without having to recombine fields (the reason 24 fps progressive DV camcorders do this). The latter shows the pattern of 2 uncombed frames and 3 combed frames.
In case it isn't clear to you, pulldown flags tell the decoder how long to display each frame. There are only two possible values: 2 fields or 3 fields. So a 29.97 video can use 2:2:2:2... flags to achieve 59.94 fields per second. A 19.98 fps video can use 3:3:3:3... flags to achieve 59.94 fields per second. Any rate in between those extremes uses a mixture of 2 and 3. To get from 23.976 fps to 59.95 fps requires a 50:50 mix of 2 and 3. But it could be 2:3... or 2:2:3:3... or 2:2:2:3:3:3... etc. Different patterns give different amounts of judder.
Why don't you just post a sample with extended motion? Use DgIndex to demux a section as M2V.Last edited by jagabo; 20th Mar 2012 at 09:52.
-
Thanks,jagabo...yes, I've been reading my whole library of saved docs on interlace/telecine,de-this and de-that. Got my head swimming, sticky notes with frame numbers all over the place. But I get your message.
Okay, so I'm gonna get this 3:2 pulldown business straight in my head if it kills me. The part I don't understand yet is this (I will post samples shortly so you can see what I'm struggling with): If I look at the original source video, it appears progressive and telecined but no combing. The same is true of both the digital cable version and the VHS capture - both sources appear to be from the same master. I can't use the VHS version of the scenes I'm dealing with here, because a JVC machine made a meal of the opening VHS titles. So I'm working with the broadcast copy I recorded at high bitrate to DVD.
After IVTC, I see only progressive frames in both sources. Except for VHS damage and horrible color on tape, they both look alike in other respects. If I use SeparateFields on either of the two, still no combing. I just did a 3:2 pulldown on the IVTC'd version, with TMPGEnc Plus and with HCenc. Both were set for 24.97 > 29.97 (3:2 pulldown). Neither encode looks exactly alike as far as combing/blending goes, but each encode does show the same frames as progressive (a/a, b/b/, e/e, f/f, h/h/ and i/i are the same in both). What is different are the interlaced frames; TMPGEnc Plus has more interlacing than HCenc. You're right -- I understand what you mean by the way they display in VirtualDub and VdubMod.
Will post a few seconds of the originals soon in m2v. Thanks for taking a look. -
Samples. Now working on the 2-minute opening titles from a broadcast copy recorded to DVD. The VHS titles were chewed up royally by a JVC player. Purchased two other VHS copies, both in horrible shape. Telecine in the broadcast copy and the VHS look exactly the same.
Broadcast_original.m2v (from a .d2v of the DVD recording off digital cable)
(MPEG/DVD, 29.97 fps, 4-seconds, 3.1 MB)
http://dc451.4shared.com/download/UbWCMXKg/Broadcast_original.m2v
VHS_Capture (AVI, 29.97fps, 640x480 YUY2/Lagarith, 1.5-seconds, 9.1 MB)
http://dc379.4shared.com/download/Iy9nibb_/VHS_capture.avi
Telecine occurs on slightly different frames because the VHS movie "begins" at a different point than the broadcast copy does. Telecine pattern is the same.
After the opening titles, I'll be using this capture for the rest of the movie. The broadcast version in several areas has worse problems than the tape (lots of smeared motion, and the same spots!), and has station logos, banners, popups.
The broadcast version was converted to AVI using DGIndex. Cleaned up with:
SmoothLevels()
Checkmate()
daa()
IVTC (TFM + TDecimate)
RemoveSpots()
light LSFMod
NeatVideo (temporal mode only)
gradation curves
The following were encoded from the resulting AVI:
Broadcast_Pulldown_HCenc.mpg
3:2 pulldown in HCenc (MPEG2, 29.97 fps, 4-seconds, 3.2 MB)
http://dc523.4shared.com/download/LNxzf3tE/Broadcast_Pulldown_HCenc.mpg
Broadcast_pulldown_TMPGenc.mpg
3:2 pulldown in TMPGenc Plus (MPEG2, 29.97 fps, 4-seconds, 3 MB)
http://dc471.4shared.com/download/MYPDL7uM/Broadcast_pulldown_TMPGenc.mpgLast edited by sanlyn; 20th Mar 2012 at 12:44.
-
Last edited by jagabo; 20th Mar 2012 at 13:06.
-
I know. That's the problem. The attached AVI above displays that, too. Tried yadif, tried QTGMC on the IVTC'd file. They won't go away. Same for the VHS capture. There seems to be some interlace effects that won't go away ? ?
Last edited by sanlyn; 20th Mar 2012 at 12:58.
-
Maybe could try different settings in TFM or TDecimate ? I used:
TFM(order=1)
TDecimate()
Spent half a day last week, reading docs and posts on IVTC, but nothing describes this.Last edited by sanlyn; 20th Mar 2012 at 13:00.
-
No, the AVI does not have frame blending.
Code:AviSource("VHS_capture.avi") AssumeTFF() SeparateFields() SelectEvery(10,0,1,3,4,6,7,8,9) Weave()
-
I don't see any field blends in "Broadcast_original.m2v" with simple IVTC ?
MPEG2Source()
TFM()
TDecimate()
So it must be some filter or encode setting used after causing that ? -
I don't think HcEnc can even do that. And since both encoders show the same problem it's likely something in your filtering. Open the AVS script in VirtualDub. Do you see the blending? What's the frame rate? When you built your D2V index was DgIndex in Honor Pulldown Flags" mode? You don't want "Forced Film" mode.
-
jagabo, you refer to the avs script you posted just above?
Ed: Using the script posted from #133 above, no. No blending.Last edited by sanlyn; 20th Mar 2012 at 13:25.
-
Well, shucks, paint me blue and call me a druid. I figured it out.
Took me a bit to make a new d2v, the mpg and old d2v were moved. The original script used daa() for anti-alias. daa() uses nneedi2 and other stuff, does some resizing, whatnot. With daa() in the script, I get blending. Remove daa(), no blend!
I'll have to smooth edges another way . The old avs was in two steps, here's the combined script which I just now ran in its entirety from scratch, with daa() commented-out. Removing daa() removed blending. Supposed to run that function before IVTC, but it does some damage here.
Well, I'll be damned -- mm, no, I'm long past damning. Another embarrassing learning experience, friends. Hope it helps others. Will clean up this combined script later, I know it looks clumsy but I was in a rush.
Code:Vv=MPEG2Source("E:\LIL2\LIL13_56k\TitleE\LIL56k_TitleApcm_E.d2v") Av=WavSource("E:\LIL2\LIL13_56k\TitleE\LIL56k_TitleApcm.wav") AV=AudioDub(Vv, Av) # == # cut start at frame 2212 - 15 frames before red spot during fade from black # cut end of resulting clip at 610 # == AVX=AV.Trim(2212,2821) AVX SmoothLevels(0, 1.0, 255, 16, 235, Limiter=0, TVrange = false, Lmode=1,\ brightSTR=5,DarkSTR=10) Checkmate(24) #daa() <-- troublemaker commented-out. Happiness reigns supreme. in2=last in2 TFM(in2,order=1) TDecimate() LSFMod(strength=25) ## -- ## -- No conversion: save as YV12 using Fast Recompress -- ## -- #Anti-aliasing with contra-sharpening by Didée function daa(clip c) { nn = c.nnedi2(field=-2) dbl = mt_average(selecteven(nn),selectodd(nn),U=3,V=3) dblD = mt_makediff(c,dbl,U=3,V=3) shrpD = mt_makediff(dbl,dbl.removegrain((width(c)>1100) ? 20 : 11),U=3,V=3) DD = shrpD.repair(dblD,13) return dbl.mt_adddiff(DD,U=3,V=3) }
Last edited by sanlyn; 20th Mar 2012 at 15:07.
-
I haven't read this thread, but I don't think you can run daa on interlaced or telecined video - it expect progressive input. It should have gone after TDecimate.
Hope this helps.
Cheers,
David. -
Last edited by jagabo; 21st Mar 2012 at 08:23.
-
I read on some anime site that anti-alias and dot crawl filters like daa() and others should be run before ivtc/delinterlace. I always questioned that (why would it be better that way, asked little ole me). Anyway, I moved daa() after TIVTC. Worked fine. It smoothed some half-broken edges and jaggies that apparently came with the cable broadcast. Odd, it doesn't appear on the VHS.
Working on re-encoding now. Busy last nite with a PC repair client -- which paid for the brand new Hauppauge HD PVR I just picked up an hour ago. Finally. -
FWIW Some clever dot crawl filters do have to come first. That part was true.
Cheers,
David. -
Yes, dot crawl on digital cable. Nothing about cable surprises me any more.
Links to clips from the re-do, same 1.5-sec and 4-sec scenes as earlier. Same avs, this time I remembered to add RemoveSpots(). Quit a difference. TMPGenc telecines 3 frames in 5, HCenc does 2 in 5. I still have to tweak some plugins, the mpg's look over filtered to me.
YV12 23.97 fps AVI (Lagarith) TIVTC'd, 7.2MB: http://dc394.4shared.com/download/7Iu_N34M/Broadcast_ivtc_new_2.avi
MPEG2/DVD 3:2 pulldown, HCenc 3MB: http://dc502.4shared.com/download/-imwxH-c/Broadcast_HCenc32_new.mpg
MPEG2/DVD 3:2 pulldown TMPGenc 3MB: http://dc238.4shared.com/download/f5z42eN6/Broadcast_TMPGenc32_new.mpg -
Hmm. Don't know where it's coming from. I'm checking settings . . .
Last edited by sanlyn; 21st Mar 2012 at 19:25.
-
I'm not talking about comb artifacts. If you look at individual frames with VirtualDubMod, or fields after a simple bob, you will see frames/fields are blends from two different film frames.
-
-
Yes, you're supposed to use the "3:2 pulldown when playback" option. The other one is a hard telecine -- which would require you to set the Encode Mode to Interlace and the frame rate to 29.97 to get a proper file. I see that TMPGEnc generates the blended frames/fields when you use the hard telecine option with the wrong encode mode and frame rate options.
Your earlier AVI encoded with TMPGEnc with both soft and hard telecine:Last edited by jagabo; 21st Mar 2012 at 20:46.
-
If I set "3:2 pulldown when playback", according to the doc, the frames will be interlaced and the fps automatically gets populated with "23.976 fps (internally 29.97 fps)" (unquote). If I select the "3:2 pulldown" button, TMPGenc fills in the frame rate as 29.97 fps (the other "internal" choice does not appear in the list). The encode mode choices then become "interlaced" or "non-interlaced".
The earlier encode used 3:2 pulldown on playback with the 23.976 fps+29.97 internal option. That gave me 3 blended/interlaced frames for every 5. If I try to choose 29.97 fps with "3:2 pulldown on playback", TMPGenc will not allow me to set the frame rate to 29.97. The only frame rates allowed with that option are 23.976/29.97 or 24/30.
For the encode posted above, I used the 3:2 pulldown box and set encode mode to non-interlaced @ 29.97 fps (TMPGenc filled-in that fps when I checked the 3:2 box).
Similar Threads
-
Set "Output filename" As Default Global "File/segment title" In MkvMerge?
By LouieChuckyMerry in forum Video ConversionReplies: 0Last Post: 9th Jul 2011, 02:52 -
Mini DV + S-VHS Combo Decks: "SR" Pro Series vs "HR" Consumer Series
By jbd5010 in forum CapturingReplies: 0Last Post: 12th Apr 2011, 23:34 -
WMV files: Changing "Recorded Date", "Media Created" fields in metadata
By axhack in forum EditingReplies: 5Last Post: 18th Sep 2010, 02:27 -
"stretching" or "cramming" aspect ratio when I convert .avi to MP4
By Agent Bauer in forum ffmpegX general discussionReplies: 3Last Post: 21st Mar 2010, 11:30 -
JVC S-VHS machine w/ stabilizer "on" plus inline TBC = disaster?
By justin81 in forum RestorationReplies: 8Last Post: 7th Oct 2009, 01:23